Quantizing autoencoders in a neural network

The performance of a neural network is improved by applying quantization to data at various points in the network. In an embodiment, a neural network includes two paths. A quantization is applied to each path, such that when an output from each path is combined, further quantization is not required....

Ausführliche Beschreibung

Gespeichert in:

Bibliographische Detailangaben
Hauptverfasser:	Munkberg, Jacob, Hasselgren, Jon
Format:	Patent
Sprache:	eng
Schlagworte:	CALCULATING COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS COMPUTING CONTROL OR REGULATING SYSTEMS IN GENERAL CONTROLLING COUNTING FUNCTIONAL ELEMENTS OF SUCH SYSTEMS MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS ORELEMENTS PHYSICS REGULATING SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
Online-Zugang:	Volltext bestellen
Tags:	Tag hinzufügen Keine Tags, Fügen Sie den ersten Tag hinzu!

container_end_page
container_issue
container_start_page
container_title
container_volume
creator	Munkberg, Jacob Hasselgren, Jon
description	The performance of a neural network is improved by applying quantization to data at various points in the network. In an embodiment, a neural network includes two paths. A quantization is applied to each path, such that when an output from each path is combined, further quantization is not required. In an embodiment, the neural network is an autoencoder that includes at least one skip connection. In an embodiment, the system determines a set of quantization parameters based on the characteristics of the data in the primary path and in the skip connection, such that both network paths produce output data in the same fixed point format. As a result, the data from both network paths can be combined without requiring an additional quantization.
format	Patent
fullrecord	<record><control><sourceid>epo_EVB</sourceid><recordid>TN_cdi_epo_espacenet_US11977388B2</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>US11977388B2</sourcerecordid><originalsourceid>FETCH-epo_espacenet_US11977388B23</originalsourceid><addsrcrecordid>eNrjZNAOLE3MK8msysxLV0gsLclPzUvOT0ktKlbIzFNIVMhLLS1KzAFSJeX5Rdk8DKxpiTnFqbxQmptB0c01xNlDN7UgPz61uCAxORWoMj402NDQ0tzc2MLCyciYGDUAjhYp_Q</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>patent</recordtype></control><display><type>patent</type><title>Quantizing autoencoders in a neural network</title><source>esp@cenet</source><creator>Munkberg, Jacob ; Hasselgren, Jon</creator><creatorcontrib>Munkberg, Jacob ; Hasselgren, Jon</creatorcontrib><description>The performance of a neural network is improved by applying quantization to data at various points in the network. In an embodiment, a neural network includes two paths. A quantization is applied to each path, such that when an output from each path is combined, further quantization is not required. In an embodiment, the neural network is an autoencoder that includes at least one skip connection. In an embodiment, the system determines a set of quantization parameters based on the characteristics of the data in the primary path and in the skip connection, such that both network paths produce output data in the same fixed point format. As a result, the data from both network paths can be combined without requiring an additional quantization.</description><language>eng</language><subject>CALCULATING ; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS ; COMPUTING ; CONTROL OR REGULATING SYSTEMS IN GENERAL ; CONTROLLING ; COUNTING ; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS ; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS ORELEMENTS ; PHYSICS ; REGULATING ; SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES</subject><creationdate>2024</creationdate><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktohtml>$$Uhttps://worldwide.espacenet.com/publicationDetails/biblio?FT=D&date=20240507&DB=EPODOC&CC=US&NR=11977388B2$$EHTML$$P50$$Gepo$$Hfree_for_read</linktohtml><link.rule.ids>230,308,776,881,25543,76293</link.rule.ids><linktorsrc>$$Uhttps://worldwide.espacenet.com/publicationDetails/biblio?FT=D&date=20240507&DB=EPODOC&CC=US&NR=11977388B2$$EView_record_in_European_Patent_Office$$FView_record_in_$$GEuropean_Patent_Office$$Hfree_for_read</linktorsrc></links><search><creatorcontrib>Munkberg, Jacob</creatorcontrib><creatorcontrib>Hasselgren, Jon</creatorcontrib><title>Quantizing autoencoders in a neural network</title><description>The performance of a neural network is improved by applying quantization to data at various points in the network. In an embodiment, a neural network includes two paths. A quantization is applied to each path, such that when an output from each path is combined, further quantization is not required. In an embodiment, the neural network is an autoencoder that includes at least one skip connection. In an embodiment, the system determines a set of quantization parameters based on the characteristics of the data in the primary path and in the skip connection, such that both network paths produce output data in the same fixed point format. As a result, the data from both network paths can be combined without requiring an additional quantization.</description><subject>CALCULATING</subject><subject>COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS</subject><subject>COMPUTING</subject><subject>CONTROL OR REGULATING SYSTEMS IN GENERAL</subject><subject>CONTROLLING</subject><subject>COUNTING</subject><subject>FUNCTIONAL ELEMENTS OF SUCH SYSTEMS</subject><subject>MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS ORELEMENTS</subject><subject>PHYSICS</subject><subject>REGULATING</subject><subject>SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES</subject><fulltext>true</fulltext><rsrctype>patent</rsrctype><creationdate>2024</creationdate><recordtype>patent</recordtype><sourceid>EVB</sourceid><recordid>eNrjZNAOLE3MK8msysxLV0gsLclPzUvOT0ktKlbIzFNIVMhLLS1KzAFSJeX5Rdk8DKxpiTnFqbxQmptB0c01xNlDN7UgPz61uCAxORWoMj402NDQ0tzc2MLCyciYGDUAjhYp_Q</recordid><startdate>20240507</startdate><enddate>20240507</enddate><creator>Munkberg, Jacob</creator><creator>Hasselgren, Jon</creator><scope>EVB</scope></search><sort><creationdate>20240507</creationdate><title>Quantizing autoencoders in a neural network</title><author>Munkberg, Jacob ; Hasselgren, Jon</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-epo_espacenet_US11977388B23</frbrgroupid><rsrctype>patents</rsrctype><prefilter>patents</prefilter><language>eng</language><creationdate>2024</creationdate><topic>CALCULATING</topic><topic>COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS</topic><topic>COMPUTING</topic><topic>CONTROL OR REGULATING SYSTEMS IN GENERAL</topic><topic>CONTROLLING</topic><topic>COUNTING</topic><topic>FUNCTIONAL ELEMENTS OF SUCH SYSTEMS</topic><topic>MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS ORELEMENTS</topic><topic>PHYSICS</topic><topic>REGULATING</topic><topic>SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES</topic><toplevel>online_resources</toplevel><creatorcontrib>Munkberg, Jacob</creatorcontrib><creatorcontrib>Hasselgren, Jon</creatorcontrib><collection>esp@cenet</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Munkberg, Jacob</au><au>Hasselgren, Jon</au><format>patent</format><genre>patent</genre><ristype>GEN</ristype><title>Quantizing autoencoders in a neural network</title><date>2024-05-07</date><risdate>2024</risdate><abstract>The performance of a neural network is improved by applying quantization to data at various points in the network. In an embodiment, a neural network includes two paths. A quantization is applied to each path, such that when an output from each path is combined, further quantization is not required. In an embodiment, the neural network is an autoencoder that includes at least one skip connection. In an embodiment, the system determines a set of quantization parameters based on the characteristics of the data in the primary path and in the skip connection, such that both network paths produce output data in the same fixed point format. As a result, the data from both network paths can be combined without requiring an additional quantization.</abstract><oa>free_for_read</oa></addata></record>
fulltext	fulltext_linktorsrc
identifier
ispartof
issn
language	eng
recordid	cdi_epo_espacenet_US11977388B2
source	esp@cenet
subjects	CALCULATING COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS COMPUTING CONTROL OR REGULATING SYSTEMS IN GENERAL CONTROLLING COUNTING FUNCTIONAL ELEMENTS OF SUCH SYSTEMS MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS ORELEMENTS PHYSICS REGULATING SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
title	Quantizing autoencoders in a neural network
url	https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-23T12%3A52%3A58IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-epo_EVB&rft_val_fmt=info:ofi/fmt:kev:mtx:patent&rft.genre=patent&rft.au=Munkberg,%20Jacob&rft.date=2024-05-07&rft_id=info:doi/&rft_dat=%3Cepo_EVB%3EUS11977388B2%3C/epo_EVB%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_id=info:pmid/&rfr_iscdi=true